tcp_in_quickack_mode() is called from input path for small packets.
It calls __sk_dst_get() which reads sk->sk_dst_cache which has been
put in sock_read_tx group (for good reasons).
Then dst_metric(dst, RTAX_QUICKACK) also needs extra cache line misses.
Cache RTAX_QUICKACK in icsk->icsk_ack.dst_quick_ack to no longer pull
these cache lines for the cases a delayed ACK is scheduled.
After this patch TCP receive path does not longer access sock_read_tx
group.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250312083907.1931644-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As mentioned in commit 648700f76b ("inet: frags:
use rhashtables for reassembly units"):
A followup patch will even remove the refcount hold/release
left from prior implementation and save a couple of atomic
operations.
This patch implements this idea, seven years later.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250312082250.1803501-5-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
In the following patch, we no longer assume inet_frag_kill()
callers own a reference.
Consuming two refcounts from inet_frag_kill() would lead in UAF.
Propagate the pointer to the refs that will be consumed later
by the final inet_frag_putn() call.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250312082250.1803501-4-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
inet_frag_putn() can release multiple references
in one step.
Use it in inet_frags_free_cb().
Replace inet_frag_put(X) with inet_frag_putn(X, 1)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250312082250.1803501-2-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
It's quite common to have a single UDP tunnel type active in the
whole system. In such a case we can replace the indirect call for
the UDP tunnel GRO callback with a static call.
Add the related accounting in the control path and switch to static
call when possible. To keep the code simple use a static array for
the registered tunnel types, and size such array based on the kernel
config.
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/6fd1f9c7651151493ecab174e7b8386a1534170d.1741718157.git.pabeni@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Most UDP tunnels bind a socket to a local port, with ANY address, no
peer and no interface index specified.
Additionally it's quite common to have a single tunnel device per
namespace.
Track in each namespace the UDP tunnel socket respecting the above.
When only a single one is present, store a reference in the netns.
When such reference is not NULL, UDP tunnel GRO lookup just need to
match the incoming packet destination port vs the socket local port.
The tunnel socket never sets the reuse[port] flag[s]. When bound to no
address and interface, no other socket can exist in the same netns
matching the specified local port.
Matching packets with non-local destination addresses will be
aggregated, and eventually segmented as needed - no behavior changes
intended.
Note that the UDP tunnel socket reference is stored into struct
netns_ipv4 for both IPv4 and IPv6 tunnels. That is intentional to keep
all the fastpath-related netns fields in the same struct and allow
cacheline-based optimization. Currently both the IPv4 and IPv6 socket
pointer share the same cacheline as the `udp_table` field.
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/4d5c319c4471161829f50cb8436841de81a5edae.1741718157.git.pabeni@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Accurate ECN needs to send custom flags to handle IP-ECN
field reflection during handshake.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ECN bits in TOS are always cleared when sending in ACKs in TW. Clearing
them is problematic for TCP flows that used Accurate ECN because ECN bits
decide which service queue the packet is placed into (L4S vs Classic).
Effectively, TW ACKs are always downgraded from L4S to Classic queue
which might impact, e.g., delay the ACK will experience on the path
compared with the other packets of the flow.
Change the TW ACK sending code to differentiate:
- In tcp_v4_send_reset(), commit ba9e04a7dd ("ip: fix tos reflection
in ack and reset packets") cleans ECN bits for TW reset and this is
not affected.
- In tcp_v4_timewait_ack(), ECN bits for all TW ACKs are cleaned. But now
only ECN bits of ACKs for oow data or paws_reject are cleaned, and ECN
bits of other ACKs will not be cleaned.
- In tcp_v4_reqsk_send_ack(), commit 66b13d99d9 ("ipv4: tcp: fix TOS
value in ACK messages sent from TIME_WAIT") did not clean ECN bits of
ACKs for oow data or paws_reject. But now the ECN bits rae cleaned for
these ACKs.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
AE flag needs to be preserved for AccECN.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are important differences in how the CWR field behaves
in RFC3168 and AccECN. With AccECN, CWR flag is part of the
ACE counter and its changes are important so adjust the flags
changed mask accordingly.
Also, if CWR is there, set the Accurate ECN GSO flag to avoid
corrupting CWR flag somewhere.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Handling the CWR flag differs between RFC 3168 ECN and AccECN.
With RFC 3168 ECN aware TSO (NETIF_F_TSO_ECN) CWR flag is cleared
starting from 2nd segment which is incompatible how AccECN handles
the CWR flag. Such super-segments are indicated by SKB_GSO_TCP_ECN.
With AccECN, CWR flag (or more accurately, the ACE field that also
includes ECE & AE flags) changes only when new packet(s) with CE
mark arrives so the flag should not be changed within a super-skb.
The new skb/feature flags are necessary to prevent such TSO engines
corrupting AccECN ACE counters by clearing the CWR flag (if the
CWR handling feature cannot be turned off).
If NIC is completely unaware of RFC3168 ECN (doesn't support
NETIF_F_TSO_ECN) or its TSO engine can be set to not touch CWR flag
despite supporting also NETIF_F_TSO_ECN, TSO could be safely used
with AccECN on such NIC. This should be evaluated per NIC basis
(not done in this patch series for any NICs).
For the cases, where TSO cannot keep its hands off the CWR flag,
a GSO fallback is provided by this patch.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Create helpers for TCP ECN modes. No functional changes.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename tcp_ecn_check_ce to tcp_data_ecn_check as it is
called only for data segments, not for ACKs (with AccECN,
also ACKs may get ECN bits).
The extra "layer" in tcp_ecn_check_ce() function just
checks for ECN being enabled, that can be moved into
tcp_ecn_field_check rather than having the __ variant.
No functional changes.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With AccECN, there's one additional TCP flag to be used (AE)
and ACE field that overloads the definition of AE, CWR, and
ECE flags. As tcp_flags was previously only 1 byte, the
byte-order stuff needs to be added to it's handling.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever timestamp advances, it declares progress which
can be used by the other parts of the stack to decide that
the ACK is the most recent one seen so far.
AccECN will use this flag when deciding whether to use the
ACK to update AccECN state or not.
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- Move tcp_count_delivered() earlier and split tcp_count_delivered_ce()
out of it
- Move tcp_in_ack_event() later
- While at it, remove the inline from tcp_in_ack_event() and let
the compiler to decide
Accurate ECN's heuristics does not know if there is going
to be ACE field based CE counter increase or not until after
rtx queue has been processed. Only then the number of ACKed
bytes/pkts is available. As CE or not affects presence of
FLAG_ECE, that information for tcp_in_ack_event is not yet
available in the old location of the call to tcp_in_ack_event().
Signed-off-by: Ilpo Järvinen <ij@kernel.org>
Signed-off-by: Chia-Yu Chang <chia-yu.chang@nokia-bell-labs.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the more esoteric helpers for netdev instance lock to
a dedicated header. This avoids growing netdevice.h to infinity
and makes rebuilding the kernel much faster (after touching
the header with the helpers).
The main netdev_lock() / netdev_unlock() functions are used
in static inlines in netdevice.h and will probably be used
most commonly, so keep them in netdevice.h.
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Link: https://patch.msgid.link/20250307183006.2312761-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
SKB_DROP_REASON_UDP_CSUM can be used in four locations
when dropping a packet because of a wrong UDP checksum.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20250307102002.2095238-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When introduced in commit 61723b3932 ("tcp: ulp: add functions to dump
ulp-specific information"), the whole ULP diag info has been exported
only if the requester had CAP_NET_ADMIN.
It looks like not everything is sensitive, and some info can be exported
to all users in order to ease the debugging from the userspace side
without requiring additional capabilities. Each layer should then decide
what can be exposed to everybody. The 'net_admin' boolean is then passed
to the different layers.
On kTLS side, it looks like there is nothing sensitive there: version,
cipher type, tx/rx user config type, plus some flags. So, only some
metadata about the configuration, no cryptographic info like keys, etc.
Then, everything can be exported to all users.
On MPTCP side, that's different. The MPTCP-related sequence numbers per
subflow should certainly not be exposed to everybody. For example, the
DSS mapping and ssn_offset would give all users on the system access to
narrow ranges of values for the subflow TCP sequence numbers and
MPTCP-level DSNs, and then ease packet injection. The TCP diag interface
doesn't expose the TCP sequence numbers for TCP sockets, so best to do
the same here. The rest -- token, IDs, flags -- can be exported to
everybody.
Acked-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20250306-net-next-tcp-ulp-diag-net-admin-v1-2-06afdd860fc9@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Since its introduction in commit 61723b3932 ("tcp: ulp: add functions
to dump ulp-specific information"), the ULP diag info have been exported
only if the requester had CAP_NET_ADMIN.
At least the ULP name can be exported without CAP_NET_ADMIN. This will
already help identifying which layer is being used, e.g. which TCP
connections are in fact MPTCP subflow.
Acked-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20250306-net-next-tcp-ulp-diag-net-admin-v1-1-06afdd860fc9@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A recent cleanup changed the behaviour of tcp_set_window_clamp(). This
looks unintentional, and affects MPTCP selftests, e.g. some tests
re-establishing a connection after a disconnect are now unstable.
Before the cleanup, this operation was done:
new_rcv_ssthresh = min(tp->rcv_wnd, new_window_clamp);
tp->rcv_ssthresh = max(new_rcv_ssthresh, tp->rcv_ssthresh);
The cleanup used the 'clamp' macro which takes 3 arguments -- value,
lowest, and highest -- and returns a value between the lowest and the
highest allowable values. This then assumes ...
lowest (rcv_ssthresh) <= highest (rcv_wnd)
... which doesn't seem to be always the case here according to the MPTCP
selftests, even when running them without MPTCP, but only TCP.
For example, when we have ...
rcv_wnd < rcv_ssthresh < new_rcv_ssthresh
... before the cleanup, the rcv_ssthresh was not changed, while after
the cleanup, it is lowered down to rcv_wnd (highest).
During a simple test with TCP, here are the values I observed:
new_window_clamp (val) rcv_ssthresh (lo) rcv_wnd (hi)
117760 (out) 65495 < 65536
128512 (out) 109595 > 80256 => lo > hi
1184975 (out) 328987 < 329088
113664 (out) 65483 < 65536
117760 (out) 110968 < 110976
129024 (out) 116527 > 109696 => lo > hi
Here, we can see that it is not that rare to have rcv_ssthresh (lo)
higher than rcv_wnd (hi), so having a different behaviour when the
clamp() macro is used, even without MPTCP.
Note: new_window_clamp is always out of range (rcv_ssthresh < rcv_wnd)
here, which seems to be generally the case in my tests with small
connections.
I then suggests reverting this part, not to change the behaviour.
Fixes: 863a952eb7 ("tcp: tcp_set_window_clamp() cleanup")
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/551
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Tested-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20250305-net-next-fix-tcp-win-clamp-v1-1-12afb705d34e@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In order to speedup __inet_hash_connect(), we want to ensure hash values
for <source address, port X, destination address, destination port>
are not randomly spread, but monotonically increasing.
Goal is to allow __inet_hash_connect() to derive the hash value
of a candidate 4-tuple with a single addition in the following
patch in the series.
Given :
hash_0 = inet_ehashfn(saddr, 0, daddr, dport)
hash_sport = inet_ehashfn(saddr, sport, daddr, dport)
Then (hash_sport == hash_0 + sport) for all sport values.
As far as I know, there is no security implication with this change.
After this patch, when __inet_hash_connect() has to try XXXX candidates,
the hash table buckets are contiguous and packed, allowing
a better use of cpu caches and hardware prefetchers.
Tested:
Server: ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog
Client: ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog -c -H server
Before this patch:
utime_start=0.271607
utime_end=3.847111
stime_start=18.407684
stime_end=1997.485557
num_transactions=1350742
latency_min=0.014131929
latency_max=17.895073144
latency_mean=0.505675853
latency_stddev=2.125164772
num_samples=307884
throughput=139866.80
perf top on client:
56.86% [kernel] [k] __inet6_check_established
17.96% [kernel] [k] __inet_hash_connect
13.88% [kernel] [k] inet6_ehashfn
2.52% [kernel] [k] rcu_all_qs
2.01% [kernel] [k] __cond_resched
0.41% [kernel] [k] _raw_spin_lock
After this patch:
utime_start=0.286131
utime_end=4.378886
stime_start=11.952556
stime_end=1991.655533
num_transactions=1446830
latency_min=0.001061085
latency_max=12.075275028
latency_mean=0.376375302
latency_stddev=1.361969596
num_samples=306383
throughput=151866.56
perf top:
50.01% [kernel] [k] __inet6_check_established
20.65% [kernel] [k] __inet_hash_connect
15.81% [kernel] [k] inet6_ehashfn
2.92% [kernel] [k] rcu_all_qs
2.34% [kernel] [k] __cond_resched
0.50% [kernel] [k] _raw_spin_lock
0.34% [kernel] [k] sched_balance_trigger
0.24% [kernel] [k] queued_spin_lock_slowpath
There is indeed an increase of throughput and reduction of latency.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Tested-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Link: https://patch.msgid.link/20250305034550.879255-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR (net-6.14-rc6).
Conflicts:
net/ethtool/cabletest.c
2bcf4772e4 ("net: ethtool: try to protect all callback with netdev instance lock")
637399bf7e ("net: ethtool: netlink: Allow NULL nlattrs when getting a phy_device")
No Adjacent changes.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After blamed commit rtm_to_fib_config() now calls
lwtunnel_valid_encap_type{_attr}() without RTNL held,
triggering an unlock balance in __rtnl_unlock,
as reported by syzbot [1]
IPv6 and rtm_to_nh_config() are not yet converted.
Add a temporary @rtnl_is_held parameter to lwtunnel_valid_encap_type()
and lwtunnel_valid_encap_type_attr().
While we are at it replace the two rcu_dereference()
in lwtunnel_valid_encap_type() with more appropriate
rcu_access_pointer().
[1]
syz-executor245/5836 is trying to release lock (rtnl_mutex) at:
[<ffffffff89d0e38c>] __rtnl_unlock+0x6c/0xf0 net/core/rtnetlink.c:142
but there are no more locks to release!
other info that might help us debug this:
no locks held by syz-executor245/5836.
stack backtrace:
CPU: 0 UID: 0 PID: 5836 Comm: syz-executor245 Not tainted 6.14.0-rc4-syzkaller-00873-g3424291dd242 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:94 [inline]
dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
print_unlock_imbalance_bug+0x25b/0x2d0 kernel/locking/lockdep.c:5289
__lock_release kernel/locking/lockdep.c:5518 [inline]
lock_release+0x47e/0xa30 kernel/locking/lockdep.c:5872
__mutex_unlock_slowpath+0xec/0x800 kernel/locking/mutex.c:891
__rtnl_unlock+0x6c/0xf0 net/core/rtnetlink.c:142
lwtunnel_valid_encap_type+0x38a/0x5f0 net/core/lwtunnel.c:169
lwtunnel_valid_encap_type_attr+0x113/0x270 net/core/lwtunnel.c:209
rtm_to_fib_config+0x949/0x14e0 net/ipv4/fib_frontend.c:808
inet_rtm_newroute+0xf6/0x2a0 net/ipv4/fib_frontend.c:917
rtnetlink_rcv_msg+0x791/0xcf0 net/core/rtnetlink.c:6919
netlink_rcv_skb+0x206/0x480 net/netlink/af_netlink.c:2534
netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
sock_sendmsg_nosec net/socket.c:709 [inline]
Fixes: 1dd2af7963 ("ipv4: fib: Convert RTM_NEWROUTE and RTM_DELROUTE to per-netns RTNL.")
Reported-by: syzbot+3f18ef0f7df107a3f6a0@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/67c6f87a.050a0220.38b91b.0147.GAE@google.com/T/#u
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250304125918.2763514-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When I read through the TSO codes, I found out that we probably
miss initializing the tx_flags of last seg when TSO is turned
off, which means at the following points no more timestamp
(for this last one) will be generated. There are three flags
to be handled in this patch:
1. SKBTX_HW_TSTAMP
2. SKBTX_BPF
3. SKBTX_SCHED_TSTAMP
Note that SKBTX_BPF[1] was added in 6.14.0-rc2 by commit
6b98ec7e88 ("bpf: Add BPF_SOCK_OPS_TSTAMP_SCHED_CB callback")
and only belongs to net-next branch material for now. The common
issue of the above three flags can be fixed by this single patch.
This patch initializes the tx_flags to SKBTX_ANY_TSTAMP like what
the UDP GSO does to make the newly segmented last skb inherit the
tx_flags so that requested timestamp will be generated in each
certain layer, or else that last one has zero value of tx_flags
which leads to no timestamp at all.
Fixes: 4ed2d765df ("net-timestamp: TCP timestamping")
Signed-off-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When __inet_hash_connect() has to try many 4-tuples before
finding an available one, we see a high spinlock cost from
the many spin_lock_bh(&head->lock) performed in its loop.
This patch adds an RCU lookup to avoid the spinlock cost.
check_established() gets a new @rcu_lookup argument.
First reason is to not make any changes while head->lock
is not held.
Second reason is to not make this RCU lookup a second time
after the spinlock has been acquired.
Tested:
Server:
ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog
Client:
ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog -c -H server
Before series:
utime_start=0.288582
utime_end=1.548707
stime_start=20.637138
stime_end=2002.489845
num_transactions=484453
latency_min=0.156279245
latency_max=20.922042756
latency_mean=1.546521274
latency_stddev=3.936005194
num_samples=312537
throughput=47426.00
perf top on the client:
49.54% [kernel] [k] _raw_spin_lock
25.87% [kernel] [k] _raw_spin_lock_bh
5.97% [kernel] [k] queued_spin_lock_slowpath
5.67% [kernel] [k] __inet_hash_connect
3.53% [kernel] [k] __inet6_check_established
3.48% [kernel] [k] inet6_ehashfn
0.64% [kernel] [k] rcu_all_qs
After this series:
utime_start=0.271607
utime_end=3.847111
stime_start=18.407684
stime_end=1997.485557
num_transactions=1350742
latency_min=0.014131929
latency_max=17.895073144
latency_mean=0.505675853 # Nice reduction of latency metrics
latency_stddev=2.125164772
num_samples=307884
throughput=139866.80 # 190 % increase
perf top on client:
56.86% [kernel] [k] __inet6_check_established
17.96% [kernel] [k] __inet_hash_connect
13.88% [kernel] [k] inet6_ehashfn
2.52% [kernel] [k] rcu_all_qs
2.01% [kernel] [k] __cond_resched
0.41% [kernel] [k] _raw_spin_lock
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Tested-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250302124237.3913746-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add RCU protection to inet_bind_bucket structure.
- Add rcu_head field to the structure definition.
- Use kfree_rcu() at destroy time, and remove inet_bind_bucket_destroy()
first argument.
- Use hlist_del_rcu() and hlist_add_head_rcu() methods.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250302124237.3913746-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There is no reason to call ipv6_addr_type().
Instead, use highly optimized ipv6_addr_any() and ipv6_addr_v4mapped().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250302124237.3913746-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When __inet_hash_connect() has to try many 4-tuples before
finding an available one, we see a high spinlock cost from
__inet_check_established() and/or __inet6_check_established().
This patch adds an RCU lookup to avoid the spinlock
acquisition when the 4-tuple is found in the hash table.
Note that there are still spin_lock_bh() calls in
__inet_hash_connect() to protect inet_bind_hashbucket,
this will be fixed later in this series.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Tested-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250302124237.3913746-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The name 'netns_local' is confusing. A following commit will export it via
netlink, so let's use a more explicit name.
Reported-by: Eric Dumazet <edumazet@google.com>
Suggested-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Remove one indentation level.
Use max_t() and clamp() macros.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After commit 8d52da23b6 ("tcp: Defer ts_recent changes
until req is owned"), req->ts_recent is not changed anymore.
It is set once in tcp_openreq_init(), bpf_sk_assign_tcp_reqsk()
or cookie_tcp_reqsk_alloc() before the req can be seen by other
cpus/threads.
This completes the revert of eba20811f3 ("tcp: annotate
data-races around tcp_rsk(req)->ts_recent").
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wang Hai <wanghai38@huawei.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-6-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
tcp4_check_fraglist_gro(), tcp6_check_fraglist_gro(),
udp4_gro_lookup_skb() and udp6_gro_lookup_skb()
assume RCU is held so that the net structure does not disappear.
Use dev_net_rcu() instead of dev_net() to get LOCKDEP support.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TCP uses of dev_net() are under RCU protection, change them
to dev_net_rcu() to get LOCKDEP support.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-4-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use two existing drop reasons in tcp_check_req():
- TCP_RFC7323_PAWS
- TCP_OVERWINDOW
Add two new ones:
- TCP_RFC7323_TSECR (corresponds to LINUX_MIB_TSECRREJECTED)
- TCP_LISTEN_OVERFLOW (when a listener accept queue is full)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We want to add new drop reasons for packets dropped in 3WHS in the
following patches.
tcp_rcv_state_process() has to set reason to TCP_FASTOPEN,
because tcp_check_req() will conditionally overwrite the drop_reason.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://patch.msgid.link/20250301201424.2046477-2-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We converted fib_info hash tables to per-netns one and now ready to
convert RTM_NEWROUTE and RTM_DELROUTE to per-netns RTNL.
Let's hold rtnl_net_lock() in inet_rtm_newroute() and inet_rtm_delroute().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-13-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
fib_valid_key_len() is called in the beginning of fib_table_insert()
or fib_table_delete() to check if the prefix length is valid.
fib_table_insert() and fib_table_delete() are called from 3 paths
- ip_rt_ioctl()
- inet_rtm_newroute() / inet_rtm_delroute()
- fib_magic()
In the first ioctl() path, rtentry_to_fib_config() checks the prefix
length with bad_mask(). Also, fib_magic() always passes the correct
prefix: 32 or ifa->ifa_prefixlen, which is already validated.
Let's move fib_valid_key_len() to the rtnetlink path, rtm_to_fib_config().
While at it, 2 direct returns in rtm_to_fib_config() are changed to
goto to match other places in the same function
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-12-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ioctl(SIOCADDRT/SIOCDELRT) calls ip_rt_ioctl() to add/remove a route in
the netns of the specified socket.
Let's hold rtnl_net_lock() there.
Note that rtentry_to_fib_config() can be called without rtnl_net_lock()
if we convert rtentry.dev handling to RCU later.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-11-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ip_fib_net_exit() requires RTNL and is called from fib_net_init()
and fib_net_exit_batch().
Let's hold rtnl_net_lock() before ip_fib_net_exit().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-10-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will convert RTM_NEWROUTE and RTM_DELROUTE to per-netns RTNL.
Then, we need to have per-netns hash tables for struct fib_info.
Let's allocate the hash tables per netns.
fib_info_hash, fib_info_hash_bits, and fib_info_cnt are now moved
to struct netns_ipv4 and accessed with net->ipv4.fib_XXX.
Also, the netns checks are removed from fib_find_info_nh() and
fib_find_info().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-9-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When the number of struct fib_info exceeds the hash table size in
fib_create_info(), we try to allocate a new hash table with the
doubled size.
The allocation is done in fib_create_info(), and if successful, each
struct fib_info is moved to the new hash table by fib_info_hash_move().
Let's integrate the allocation and fib_info_hash_move() as
fib_info_hash_grow() to make the following change cleaner.
While at it, fib_info_hash_grow() is placed near other hash-table-specific
functions.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-8-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will allocate the fib_info hash tables per netns.
There are 5 global variables for fib_info hash tables:
fib_info_hash, fib_info_laddrhash, fib_info_hash_size,
fib_info_hash_bits, fib_info_cnt.
However, fib_info_laddrhash and fib_info_hash_size can be
easily calculated from fib_info_hash and fib_info_hash_bits.
Let's remove fib_info_hash_size and use (1 << fib_info_hash_bits)
instead.
Now we need not pass the new hash table size to fib_info_hash_move().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-7-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will allocate the fib_info hash tables per netns.
There are 5 global variables for fib_info hash tables:
fib_info_hash, fib_info_laddrhash, fib_info_hash_size,
fib_info_hash_bits, fib_info_cnt.
However, fib_info_laddrhash and fib_info_hash_size can be
easily calculated from fib_info_hash and fib_info_hash_bits.
Let's remove the fib_info_laddrhash pointer and instead use
fib_info_hash + (1 << fib_info_hash_bits).
While at it, fib_info_laddrhash_bucket() is moved near other
hash-table-specific functions.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-6-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Every time fib_info_hashfn() returns a hash value, we fetch
&fib_info_hash[hash].
Let's return the hlist_head pointer from fib_info_hashfn() and
rename it to fib_info_hash_bucket() to match a similar function,
fib_info_laddrhash_bucket().
Note that we need to move the fib_info_hash assignment earlier in
fib_info_hash_move() to use fib_info_hash_bucket() in the for loop.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-5-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will allocate fib_info_hash[] and fib_info_laddrhash[] for each netns.
Currently, fib_info_hash[] is allocated when the first route is added.
Let's move the first allocation to a new __net_init function.
Note that we must call fib4_semantics_exit() in fib_net_exit_batch()
because ->exit() is called earlier than ->exit_batch().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-4-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Both fib_info_hash[] and fib_info_laddrhash[] are hash tables for
struct fib_info and are allocated by kvzmalloc() separately.
Let's replace the two kvzmalloc() calls with kvcalloc() to remove
the fib_info_laddrhash pointer later.
Note that fib_info_hash_alloc() allocates a new hash table based on
fib_info_hash_bits because we will remove fib_info_hash_size later.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20250228042328.96624-3-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
